On Dutch allemaal and West Ulster English all Hilda Koopman UCLA October 2009 Koopman@ucla.edu This squib compares the similar distributions of floating allemaal 1 and all under wh-movement in Dutch and West Ulster English (WUE) (McCloskey 2000). A comparison between the two languages leads to the conclusion that the floated quantifier is merged at the edge of vp in both languages. A close comparison of the orders in Dutch and WUE yield direct insights in necessary properties of the derivations of WUE and Dutch, given external and internal Merge, universal head complement order (Kayne, 1994), overt (but no covert) phrasal movement; strict locality (sisterhood, not Agree), Minimality, and some condition yielding that-t. The difference between the two languages turns out to be minimal and follows mainly [h1]from a difference in the size of phrasal pied-piping (Koster 2000, Koopman and Szabolcsi 2000). The basic problem: WUE and Dutch McCloskey (2000) shows that West Ulster English allows both the standard English pattern with all immediately following a bare wh-phrase, and a stranding pattern. All may be stranded in the clause where the wh-phrase originates, or, in case of wh-movement out of CPs, in a position following a subordinating verb and preceding that. Dutch allemaal shows a (partially) similar * This squib is based on lectures on Dutch allemaal and floating quantifiers from a comparative angle (Dutch, standard English, WUE and Malagasy) presented in various places (Paris Ealing (2007), Vienna (2007), Dharamkot (2007), and the University of Utrecht (2009)). Audiences are hereby gratefully thanked for feedback.
distribution, but importantly, it can never follow the verb in non root environments and precede dat (2c). (1) a. What did he say (that) he wanted all? McC (8b) b. Wat heeft hij gezegd dat hij allemaal wilde hebben? (2) a. What did he say all that he wanted? McC (8c) b. Wat heeft hij (allemaal) gezegd (*allemaal) dat hij wilde hebben McCloskey argues that WUE all is stranded in Spec, CP, and then uses all to probe the architecture of WUE vp/vps. Dutch (2b) shows that allemaal cannot be stranded in CP: it occurs instead at the edge of vp (it precede external arguments: wat hebben er allemaal voor mensen voor squibs geschreven what kind of people write what kinds of squibs ). This pattern holds more generally in Dutch, as Barbiers (2002) first established for other discontinuous constituents. Barbiers argues stranding in CP is excluded because wh-movement never proceeds through that CP (a conclusion I am not willing to accept.) Wh-movement must proceed through the edge of vp though, thus providing welcome empirical support for Chomsky s (1986) wh-movement-viathe-edge of vp proposal. The question is how to reconcile the analyses for WUE and Dutch all/allemaal stranding, two genetically related languages, with what looks like similar types of that complements. It seems reasonable to assume all and allemaal are basically similar. Since Dutch allemaal is visibly not in Spec, CP, but at the edge of vp, McCloskey s analysis of floating in Spec, CP does not carry over. Fortunately, analyzing WUE all as occurring at the left edge of vp, just as Dutch allemaal, is not just possible (this is entertained, but rejected, by McCloskey(2000, p62, fnt6): given certain theoretical results this is in fact the only analytical possibility. In light of his conclusion that Ds are merged outside VPs in the spine, Sportiche (1996: 230, 1999: 697) updates his 1988 base generation and stranding Q-float analysis. Floated
quantifiers are merged in the spine, where they take surface scope. They quantify over a DP that has moved through its specifier, thus capturing the syntactic distribution. (cf in particular Collins 2007). This consolidates the two dominant Q float analyses -- the adverbial (VP adjoined) analysis and Sportiche 1988. Stranded WUE all and Dutch allemaal thus occur at left edge of vp boundary, because they are merged there. All/allemaal takes a vp as its complement, and attracts its restriction (i.e. the whphrase) to its Spec. Further wh-movement leads to a stranded Q. Dutch (2b) shows that stranding in CP is not available. This should now be understood as a reflection of what all/allemaal can merge with: it selects for a vp predicate, but never for an argument CP. Give this, the following minimal properties must hold in the syntactic derivations of WUE and Dutch Q-float: (3) w(h)at i [ <w(h)at> all/allemaal [ vp v [ CP <w(h)at> th/d/at) [..<w(h)at> By systematically comparing Dutch and WUE linear orders, we can use (3) to backwards engineer further derivational properties, finding the derivational paths that capture what is identical in both languages most directly. (For reasons of space, this exercise focuses on how the orders in Dutch yield insights in the particular derivations for WUE). Let us warm up with the analysis of (2a) in WUE. Stranded all shows the vp bracket must be to the left of the CP; thus a constituent containing say must have moved to the left of all, stranding the that CP. (4) a. What i did he say [ allp t i all [ vp [ CP t i that he wanted t? b. Wat i heeft hij [ allp t i allemaal [ vp gezegd [ CP t i dat hij t wilde hebben
This shifted constituent can contain a light PP to him (5). (The? mark below refers to what McCloskey argues is deviance of the optimal prosody V(pronoun) all.? contrasts with *). (5)?What did he [say to him]all that he wanted to buy t? McC (15c) wat heeft hij allemaal [tegen hem gezegd] dat hij wilde kopen *What did he say all to him that he wanted to buy McC (15b) Say to him all in (5) mirrors Q PP V in Dutch. We thus conclude that say to him in WUE English rolls up and pied-pipes to the left of all. Since only constituents can be moved, and CPs are stranded, the constituency in (6) must hold at some point in the derivation (either by what Koster calls parallel structures or, in theories with greater derivational depth, by building these up via movement from small atomic pieces (Kayne 2000, 2003, Sportiche (1999), or in my own work): (6) [what i all [ [ vp you say t i to him] [ CP t i that.] Given the basic theoretical tools and assumptions outlined above, what other properties of WUE derivations can we infer from a simple comparison of WUE orders and Dutch orders? To pursue this basic enterprise a la Koster, we start out with a basic list of the different possible scenarios in (7) and (8). We then show all instances are actually attested in the WUE sample presented in McCloskey (2000). WUE derivations minimally involve pied-piping of vp with vp sometimes pied-piping another constituent, sometimes stranding a constituent. (Discussion of the motivation for vp is postponed to the end of this squib). (7) The WUE order V all is derived by: I. pied-piping ( roll-up ) for elements that appear between allemaal and V in Dutch.
D (w) allemaal [. V ]= WUE [ V... ] (wh) all II. pied-piping followed by vp extraction for elements that appear to the left of allemaal in Dutch: D:.. XP i.. [ (w) allemaal [ t i V = WUE [ vp V..t i.. ] XP [ vp.. t i. ] (wh) all (8) The WUE order V all XP is derived by: III. pied-piping of [[vp] (what) all] to the left of elements that merged higher than Dutch allemaal : D: YP [(w) allemaal. V] = WUE [[ vp ] [(wh)all ] YP [[ vp ][ all.. IV. vp movement stranding elements that must occur below all. (cf also (4a)) D (w) allemaal [ XP. V ]= WUE [ vp... ] (wh) all [ vp.. ]XP I. When the vp rolls up: [V ] all [ ] This order is observed for indirect object PPs ((2c)?..talk to him all), for stranded Ps, which must appear between allemaal. V (9), as well as for directional PPs discussed under III: (9) a.?who did you [talk to] all McC (21b) de mensen waar je allemaal mee (over Jan Vat) gepraat hebt 2 b. *Who did you [talk all] to? McC (21a) *de mensen waar je mee allemaal gepraat heb 2 Only (inanimate) R-pronouns can strand Ps. Relative R-pronouns are compatible with animate antecedent, hence the switch to Dutch relatives.
(10) a.?who did you [give tea to] all? McCl (23a) de mensen waar je allemaal [thee aan gegeven] hebt II. When vp moves repeatedly: (V XP all). Dutch weak pronouns, indirect objects (DPs), and definite direct objects must precede allemaal, as is true for WUE non subject pronouns, indirect object DPs and definite direct objects, as well; Dutch and WUE line up in the same linear orders (as shown in the underlining). Failure to shift yields ungrammaticality (11c): (11)a. Who did Frank tell you all that they were after t? (McC 9a) Wie heeft Frank je allemaal verteld dat ze achtervolgden? b.?what did he tell his friends/mickey all (that) he wanted t? Wat heeft hij zijn vriend(sg)/mickey allemaal verteld dat hij wou hebben c *What did he tell all his friends/mickey (that)he wanted t? *Wat heeft hij allemaal zijn vriend/mickey verteld dat hij wou hebben In ECM contexts, the accusative object must shift in both languages, hence the ungrammaticality of (12b) under the reading in (12a): (12)a. Who did you want your mother [who all [ vp ] to meet at the party? McC (39a). wie heeft hij je moeder [ wie allemaal [ vp op het feest laten ontmoeten Who all did he get his mother to meet at the party b. *Who did you want all your mother to meet at the party? McCl (40a) *wie heeft hij allemaal je moeder op het feest willen laten ontmoeten How exactly do pronouns (or (in)direct objects) shift to a position preceding all? Through independent movements or by never leaving the vp (Nilsen 2003, Koster 2000)? (12a) shows the
DP or pronoun in the want vp, which requires it to have extracted from the meet vp by itself. This suggests the following derivational path for WUE with vp pied-piping preceding extraction of the pronoun (as in much of Kayne s work, and in Malagasy, Koopman 2005) (13)a. [[ vp Frank [tell [you/mickey] <tell> t cp ] [what [all.. [ CP dat.. Dutch shows the IO pronoun must extract further, as adverbs come in between. This leads to the step in (13b): extraction of the pronoun to a designated position for pronouns, followed by remnant vp movement (13c), necessary to smuggle the subject over the intervening indirect object (see below). b. you/him [ vp Frank [tell <you>.. ] what all.. e. [ vp Frank tell [ you/him [t vp ] <wat> all III. WUE [vp [what all] XP This order arise by pied-piping around an XP which occurs to the left of allemaal in Dutch. This looks like a rather straightforwardly option for the position of right peripheral adverbs in WUE. (14) What did she buy all in Derry (*all) yesterday] *all]? (McC (25), (26)) As the Dutch translations show however the analysis of the WUE string is quite tricky. (15 ) a wat heeft ze allemaal (gisteren) in Derry gekocht (gisteren)? b wat heeft ze (gisteren) (in Derry) allemaal gekocht (in Derry) (gisteren) c. wat heeft ze allemaal gekocht in Derry gisteren In particular, the locative can occur inside the allemaal..v frame (15a). This indicates the locative in (14) could be stranded to the right of all, an option which must be independently available, as WUE and Dutch wh-locatives can combine with a stranded all, where did they go
all for their holidays? (McCloskey (3c) waar is hij allemaal in zijn vakantie geweest. More research is needed here, but at least for high adverbs III looks like a reasonable option. IV. V all XP orders: vp extraction, stranding XP below all. Last but not least, this order alternates with roll up for many elements that occur between allemaal and the verb in Dutch. Judging from McCloskey s examples, the availability of this order seems to depend on whether what wh-moves is (i) a non-subject which normally occurs to the left of a PP is or (ii) a subject of v. Dutch PPs that must occur between allemaal and V allow more than one possible order in WUE, suggesting the availability of both a pied-piping derivation (subject to usual heaviness) and a stranding derivation. (16) Wat heb je allemaal in de la gestopt (gisteren) (17) a. What did you put all in the drawer? (McCl 29b) b.?what did you put in the drawer all (yesterday)? (McCl 29a) We thus see vp extraction stranding PP (17a), as well as the expected vp pied-piping PP (17b) (7.I). Both derivations must therefore be available within the same grammar: this can be described as variability in the size of pied-piping. Subject/ non subject asymmetries and vp movement McCloskey shows that WUE subject wh-movement and all stranding present tough analytical problems. First, we find an illustration of (8.II): a WUE PP which occur in Dutch between allemaal V is forced to strand, even though it must roll up when a non-subject is extracted, judging from (cf (5) what..said to him all/*said all to him ) When a wh subject is extracted
however, this pied-piping is prohibited (18a), and the Dutch pattern of PP stranding obligatorily emerges ((18b) lines up perfectly with the Dutch order). (18) a. *Who was talking to the kids/to him all last night? (McC56b, c) b.?who was talking all to the kids last night? (McC56d) c. Wie waren er allemaal met de kinderen/met hem aan het praten (gisteravond) To see what new insights the current approach can provide, let us consider (5a) above, replace he with who, and say what with talk, and examine the derivational options. The surface string shows the to PP must be stranded, hence the bracketing at some point in the derivation must be as in (19): (19) [ all [[ vp who talk] [to him/to the kids]] At this point in the derivation, who must move to all to yield a stranded all. Wh-movement to Spec, all can attract wh-phrases from embedded CPs (2a), a diagnosis for the presence of a phasal boundary. That-t configurations teach us, subject wh-phrases cannot extract when they are next to a phase boundary. (19) then is a that- t configuration, and wh-subject extraction is blocked, leaving no other option than for who to pied-pipe vp to Spec, all. Pied-piping is argued to be generally forced for wh-subject extraction in English (Koopman 2000). Quantification under pied-piping is independently attested with PPs in Dutch ( met wie heb je allemaal met wie gepraat lit.with who have you all talked ) ). There are only two pied-piping options in (19): vp pied-pipes PP or it does not. vp movement of who talk converges to (18b) all to the kids, but vp pied-piping PP does not ((18a) *who was talking to him all. Is there are any reason to suppose vp pied-piping PP could be excluded in this particular context? Abels 2003 and Kayne 2005 both propose generalizations which have precisely this effect. For Abels the complement of a phase head cannot move. In Kayne s proposal a direct complement of a head may never move to its
Spec. If these generalizations are correct, pied-piping of the entire complement to Spec, all is effectively ruled out, leaving ((18b) as the only option. Stranding to him/dp then might be available only if the unmarked pied-piping option is independently blocked. In cases of nonsubject extraction this problematic situation never arises, as the wh phrase moves to Spec all by itself, and the vp shift around the allp presumable to some other functional head. (WUE) vp movement to the left of all is forced for stranding under wh-movement, subjects or not (but perhaps other derivations are allowed in non-stranding cases). vp movement however is never possible under Q-float of the subject under A-movement, which behaves in WUE just like standard English here (McCloskey p77). Why? The fact that vp shifts around all opens up a new perspective to understand this distribution. Take non subject extraction: if what all marks a phase boundary, the external argument cannot cross it under A-movement, yet it clearly undergoes A movement, as it shows up in Spec, TP (2a). Note now that vp shifting smuggles the external argument around the boundary, from where it can map to Spec, TP in the usual fashion. This suggests vp shift is necessary to help the subject get around the phasal boundary for non subjects. For subjects, pied-piping is forced to the left of all shifting the vp for a different reason. In the case of A-movement, it is widely assumed no phasal boundary intervenes, hence vp shifting is not necessary. This implies in other words that vp shifts as high as it needs to to yield convergence. It is easy to extend this to vp movement over shifted objects and PPs, this time because of a Minimality violation (generalizing Koopman, 2008). Note that if this correct, it must be the case for Dutch as well that the subject is carried over the phasal boundary by a verbal constituent that contains it: Dutch vps then move even higher than English vps (Kayne, 1994), which implies that the derivations of Dutch and English might be even more similar than Koster (2000) claims, with Dutch simply always moving a tiny sized vp.
Comparative work on the basis of extremely simple word order patterns directly leads to conclusions about English syntax (English has object shift) which used to be hotly contested. Our investigation shows that there are deep similarities between Dutch and English, which the right theoretical tools can bring to light, and can capture quite directly as due to variability in the size of pied-piping. References. Abels, Klaus 2003 Successive Cyclicity, Anti-locality, and Adposition Stranding, UConn PhDThesis. Barbiers, Sjef 2002 Remnant Stranding and the theory of Movement, in Alexiadou et all (eds) Dimensions of Movement: from features to remnants 47-69. Collins, Christopher 2005. A smuggling approach to the passive in English. Syntax 8.2. Collins, Christopher 2007 Floated quantifiers and Linkers, ms. NYU. Kayne, Richard. 2000. Parameters and Universals. Oxford, Oxford University Press. Kayne, Richard 2005 Movement and Silence, Oxford University Press, New York. Koopman, Hilda. 2000. The Syntax of Specifiers and Heads. Routledge, New York. Koopman, H. 2005 Malagasy Imperatives in Jeffrew Heinz and Dimitrios Nthelitheos (eds) Proceedings of AFLA XII, UCLA Working Papers in Linguistics, no.12. Koopman, H. 2008. Samoan ergativity as double passivization to appear in a festschrift. Koopman, H. and A. Szabolcsi. 2000. Verbal Complexes Current Studies in Linguistics 34. Cambridge MA: The MIT Press.
Koster, Jan 2000 Pied Piping and The Word Orders of English and Dutch. In: M. Hirotani, A. Coetzee, N. Hall and J.-Y. Kim, eds., NELS 30: Proceedings of the North East Linguistic Society. GLSA, Amherst, MA, pp. 415-426. McCloskey, James. 2000. Quantifier Float and Wh-movement in an Irish English, Linguistic Inquiry, Volume 31, Number 1, 57 84. Sportiche, Dominique. 1988. A Theory of Floating Quantifiers and Its Corollaries for Constituent Structure. Linguistic Inquiry 19.3, pgs. 425-449. Sportiche, Dominique. 1996. Clitic Constructions. In Johan Rooryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon. Kluwer Academic Publishers, Dordrecht. Sportiche, Dominique. 1999. Pronominal Clitic Dependencies. In Henk van Riemsdijk (ed.), Clitics in the Languages of Europe, pgs. 679-708. Berlin: Mouton de Gruyter. Sportiche, Dominique. 2005 Division of Labor between Merge and Move: Strict Locality of Selection and Apparent Reconstruction Paradoxes, in Nathan Klinedinst et al Proceedings of the Workshop Divisions of Linguistic Labor, The La Bretesche Workshop